ISTITUTO SUPERIORE DI SANITÀold.iss.it/binary/publ/cont/08-24 web.1222954103.pdf · vascolari,...

ISSN 1123-3117 Rapporti ISTISAN

08/24

ISTITUTO SUPERIORE DI SANITÀ

Wearable device for skin contact thermography: design, construction and testing

Daniele Giansanti, Giovanni Maccioni

Dipartimento di Tecnologie e Salute

Presidente dell’Istituto Superiore di Sanità e Direttore responsabile: Enrico Garaci Registro della Stampa - Tribunale di Roma n. 131/88 del 1° marzo 1988 Redazione: Paola De Castro, Sara Modigliani e Sandra Salinetti La responsabilità dei dati scientifici e tecnici è dei singoli autori. © Istituto Superiore di Sanità 2008

Istituto Superiore di Sanità Wearable device for skin contact thermography: design, construction and testing. Daniele Giansanti, Giovanni Maccioni 2008, 28 p. Rapporti ISTISAN 08/24

The need for wearable devices for thermal monitoring is rising. These devices could be used to continuously monitor patients for breast cancer investigation or vascular, dermatological and rheumatic disorders, in viability studies, or during physical exercise. We designed and constructed a wearable device for skin-contact thermography that uses integrated silicon sensors. The device was validated using a phantom with a dynamic bench test. The thermal resolution was greater than 0.03°C, and the spatial resolution was equal to 1.6×10-5 m2. We also investigated the device’s performance on five clinical subjects. Results of these studies showed a maximal error of less than 0.1°C in each evaluation.

Key words: Wearable Device, Thermography, Contact thermography Istituto Superiore di Sanità Dispositivo indossabile per la termografia a contatto della pelle: progetto, costruzione e validazione. Daniele Giansanti, Giovanni Maccioni 2008, 28 p. Rapporti ISTISAN 08/24 (in inglese)

È oggi pressante la richiesta di dispositivi indossabili per il monitoraggio termico della pelle. Questi dispositivi potrebbero essere utilizzati per il monitoraggio continuo del tumore alla mammella o per il monitoraggio di disordini vascolari, dermatologici, circolatori e durante l’esercizio fisico. È stato progettato e testato un dispositivo indossabile per la termografia a contatto della pelle. Le prove su banco di test per la validazione sono state realizzate utilizzando anche un fantoccio progettato appositamente. La risoluzione termica è risultata migliore di 0,03°C e la risoluzione spaziale uguale a 1,6×10-5 m2. Il dispositivo è stato inoltre testato su 5 soggetti in un’applicazione medica appositamente disegnata. I risultati hanno dimostrato un errore massimo sempre inferiore a 0,1°C in valore medio.

Parole chiave: Dispositivo indossabile, Termografia, Termografia a contatto Per informazioni su questo documento scrivere a: [email protected]; [email protected] This work was funded by the internal triennial research plane Piano triennale di ricerca “Sviluppo di strumenti indossabili innovativi per il monitoraggio termografico” approved on 29/12/2005. Il rapporto è accessibile online dal sito di questo Istituto: www.iss.it. Citare questo documento come segue:

Giansanti D, Maccioni G. Wearable device for skin contact thermography: design, construction and testing. Roma: Istituto Superiore di Sanità; 2008. (Rapporti ISTISAN 08/24).

Rapporti ISTISAN 03/xxxx

TABLE OF CONTENTS

Introduction ........................................................................................................................................ 1

1. Design and construction.......................................................................................................... 2 1.1. Odoscope unit ........................................................................................................................... 3 1.2. Service unit ............................................................................................................................... 4

1.2.1. Multiplexing circuit .................................................................................................... 4 1.2.2. Processing and conditioning circuit ............................................................................ 4 1.2.3. Powering, oscillation, connecting and stabilisation circuit ......................................... 4

1.3. Elaboration unit and system integration.................................................................................... 4 1.3.1. Complete architecture of the system........................................................................... 6

1.4. Choice of the material at the interface ...................................................................................... 7 1.4.1. Desirable properties of the material ............................................................................ 7 1.4.2. Modification of the package at the interface............................................................... 7 1.4.3. Interface material ........................................................................................................ 9

2. Testing ............................................................................................................................................. 10 2.1. Design of a phantom for the bench testing................................................................................ 10

2.1.1. Generated waveforms ................................................................................................. 13 2.2. Static bench test ........................................................................................................................ 13 2.3. Phantom-based bench test ......................................................................................................... 13

3. Design of the clinical validation............................................................................................ 15 3.1. Protocol ..................................................................................................................................... 15 3.2. Outcome from the clinical validation........................................................................................ 15

3.2.1. Display of clinical data ............................................................................................... 16

4. In depth analysis ......................................................................................................................... 18 4.1. Details on the device positioning .............................................................................................. 18 4.2. Assuring aseptic conditions ...................................................................................................... 18 4.3. Device certification ................................................................................................................... 18 4.4. Device life, costs and maintenance ........................................................................................... 18

5. Discussion and conclusion .................................................................................................... 20

References .......................................................................................................................................... 22

Appendix Technical details of used component and circuits............................................................................ 25

Rapporti ISTISAN 08/24

1

INTRODUCTION

Skin temperature is influenced by complex interactions between heat conduction areas, i.e. blood flow through the various skin layers, and the external environment (1).

Thermal imaging of the skin has been used for several decades to monitor human temperature distribution. Abnormalities, such as malignancies, inflammation and infection, cause localised increases in temperature, which appear as hot spots or asymmetrical patterns on skin thermograms. Even though it is non-specific, thermography is a powerful detector of irregular patient physiology. Some successful areas for thermography use are with vascular disorders, rheumatic diseases, oncology (especially breast cancer), tissue viability, neurology, dermatological disorders, neonatal, ophthalmology, and surgery. The demand for continuous physiological monitoring via wearable instruments (2) is rising. Analysis of the literature shows the need for the development of a wearable thermographic device.

Many technologies are suitable for skin thermography. The most commonly used technologies are infrared-based, which are not suitable for wearable continuous monitoring applications. This is due to their encumbrance and complexity. The solution to this problem could come from thermal contact sensors, such as thermistors, thermocouples and/or IC (Integrated Component) sensors. Bolton (3) recently proposed a wearable system that consisted of a compact data logger for ambulatory skin temperature measurement. He illustrated an eight-channel, high-resolution thermometer system based on surface montage thermistors for the monitoring of eight different points. This work was the first to solve the problems of system transportability, data acquisition and storage and long-term analysis.

Other commonly used contact thermographic techniques in medicine are based on foils of allosteric crystals placed directly on the skin, but these cannot be used for continuous data storage. Furthermore, they suffer from the typical problems of skin contact thermography. In fact, the skin-foil interaction influences the data obtained.

The last basic aspect in contact thermography is the attachment of the sensor, which affects the heat distribution of the skin surface itself. Sweating should also be avoided in contact thermography, as it could cause a temperature drop.

We previously developed and validated a simulation environment with Matlab R13 (Mathworks, USA) and P-spice (Interlink, USA) (8). We used this environment to test different thermal contact devices based on thermocouples, thermistors, and integrated silicon components inserted on an opportunely conditioned (low-pass-filter and amplification) measurement chain (8). The device sensor surface area was varied from 16 (4×4) to 256 (16×16). The resolution, accuracy, hysteresis, and linearity were analysed. The long-term measurement stability was determined. The simulations completed in this previous study showed that the lm335 IC (National, USA) performed the best for the purpose of continuous temperature monitoring.

This technical report illustrates the development of the simulated wearable (8) system based on the selected IC with minimal encumbrance and adequate thermal and area resolution. The investigation of device performance was also based on medical application criteria with particular regards to the skin interface. The device contained a matrix of 16 sensors (4×4) for simplicity of presentation without losing scientific generality.


1. DESIGN AND CONSTRUCTION

The study implied: the design and construction of the wearable device. This device consisted of:

– Odoscope unit including 4×4 Silicon Kelvin precise (K-Sr0–K-Sr15) sensor thermometers;

– Service unit including a multiplexing circuit, a processing and conditioning circuit, a powering oscillation, connecting and stabilisation circuit;

– Elaboration unit and system integration including a personal computer, software for the analysis and connecting interface.

In order to optimize the interface of the wearable, different composite elastomeric materials have been investigated. This should minimise limitations common in contact thermography.

Figure 1 and 2 show the core elements of the system respectively the hardware of the wearable device and the constructed odoscope before the affixation the “interface sponge” and the constructed service unit.

A/Dconverter

AnalogMux

396MaximUSA

AN0

µP Pic16F877microchip

Condition chain

K-SR0

K-SR15

NO1

NO16

G0=83.34 CF=1000Hz

Out

WTS unit Service unitOdoscopeunit

Gc=83.34 CF=624 Hz

A/Dconverter

AnalogMux

396MaximUSA

AN0


Condition chain

K-SR0

K-SR15

NO1

NO16

G0=83.34 CF=1000Hz

Out

WTS unit Service unit

A/Dconverter

AnalogMux

396MaximUSA

AN0


Condition chain

K-SR0

K-SR15

NO1

NO16

G0=83.34 CF=1000Hz

Out

WTS unit Service unitOdoscopeunit

Gc=83.34 CF=624 Hz

Figure 1. Hardware of the wearable device

2


3

a

b

Figure 2. Designed and constructed odoscope before the affixation the “interface sponge” (a) and designed and constructed service unit (b)

1.1. Odoscope unit

This unit has been optimized to monitor skin temperature in the ranges of variation of human body as enlightened by Zontak (9).

The thermal sensor unit was arranged in a 4 rows by 4 columns matrix box. Each cell of the matrix box corresponded to an area of 4 mm × 4 mm. The box was monitored by one silicon Kelvin sensor thermometer-based lm335 component (10). The face of the matrix box was put into contact with the skin. The sensing head was made of a transparent sponge, described later, and carried the 16 thermal sensor packages (model lm335, National, USA). The core of the circuit was the lm335, chosen for its optimal calibrated error response. It performed in the range of interest (see Appendix) (7, 10). We preferred to use the thermal sensor lm335 in the Kelvin configuration because the circuital polarisation was easy (only a resistance was connected both to the lm335 and to the power supply). It did not need trimmers and/or potentiometers, which could cause instabilities and deviations with time. These often occurred with the Celsius configuration (10) (see Appendix). The conversion into Celsius scale can be easily completed with software.


1.2. Service unit

1.2.1. Multiplexing circuit

Thermal variations in clinical applications were not rapid. Therefore, the selection time of each channel was not a critical parameter. We used a 16:1 multiplex. It consisted of one 396 (MAXIM, USA) component, which selected one of the 16 sensors in the Time Division (TD).

1.2.2. Processing and conditioning circuit

The conditioning chain read a non-inverting amplifier designed with the lm308 operational and a low-pass filter developed with a second order Sallen & Key cell in the Butterworth configuration. In order to obtain full dynamics at the output, the Cascade Gain Gc was set at 83.34.

The cut-off frequency was set to 624 Hz, which was the value obtained through simulations. A frequency swap also confirmed the optimisation values (see appendix for the other parameters of the Sallen & Key cell). A PIC (Programmable Interface Controller) 16F877 microprocessor was used for the data connection with the PC control, multiplexing unit TD control, and the data conversion with a 10 bit-A/D converter.

The connection to the PC was done by means of the serial port. The software interface was developed with Matlab R12 for the Kelvin to Celsius conversion for the real-time display and thermographic colour mapping associated with the 16 channels.

1.2.3. Powering, oscillation, connecting and stabilisation circuit

The unit is composed of one 4MHZ oscillator M.E.C.Y.Oz (MEC, USA), and one L7805 regulator (ST, USA) in the stabilisation circuit.

1.3. Elaboration unit and system integration

The integration of the previous described units was optimized in terms of area occupancy. The layout of the odoscope unit was 4×10-4 m2. The layout of the service unit was 2.5×10-3

m2. The weight of the equipment with cases and cables was 80 g for the odoscope and 270 g for the service unit.

The component part list chosen to optimize the layout contained the following optimized packages: TO-92 for the thermal sensors, dual in-line 16 pin package for the IC Amplifiers, dual in-line 28 pin package for the antilog multiplex, Mrs25 package for the resistance, and mica-silver radial condensers for high stability and performance.

The spatial resolution was pr=1.6×10-5 m2 for a field of view of 16 pixels with an area of 16× pr.

Figure 3 shows the service unit before and after boxing. Figure 4 details a totally wearable modified version of the wearable system (odoscope and service unit) specifically designed for the clinical application described in section 3.1.

4


5

a b

Figure 3. Service unit before boxing the case (a) and after boxing the case (b)

a

b

Figure 4. Details a the totally wearable modified version of the wearable system (odoscope and service unit) specifically designed for the clinical application


1.3.1. Complete architecture of the system

Figure 5 shows the complete architecture of the two wearable systems: 1) the first architecture allows a partially wearable monitoring solution and has been

thought for the clinical applications designed in a definite volume of space (tapis-roulant; cyclo-ergometer);

2) the second architecture allows a totally wearable solution, suitable, for example, in the case of the home-daily-monitoring.

The principal differences between the two architectures are: 1) The power-supply in the totally wearable solution is based on 4 rechargeable (by three

extracell NI-MH rechargeable batteries with 3.6 V and a capacity Imax 160 mA×hour batteries, while in the partially wearable solution is directly furnished and converted by the domestic-net.

2) The connection to the PC is assured by two telemetric AUR°EL XTR-434H RX-TX modules, while in the partially wearable-solution is assured by one cable-based DAQ-PAD 6020E (National, USA) solution.

3) The case of the service unit in the totally wearable solution can be affixed at the belt-level.

O

doscopeunit

Service unit

Powersupply

12 V12 V

DAQ-PAD 6020E (A/D converter)

PC

a) PARTIALLY-WEARABLE SOLUTION

XTR-434H (Aur°el, Italy)

RX-TX

XTR-434H (Aur°el, Italy)

RX-TXOdoscopeunit

Service unit

PC

b) TOTALLY-WEARABLE SOLUTION

Figure 5. Two completed alternative architectures of the wearable system

6


7

1.4. Choice of the material at the interface

1.4.1. Desirable properties of the material

Particular care was dedicated to the choice of the material for the interface of the odoscope with the skin.

It is fundamental to respect the following requirements to minimise the typical problems of contact thermography:

1. The heat exchange, Q, between the skin and the external environment should not be hampered. This quantity can be obtained by means of the Fourier law depending on a wide range of aspects. It should last the entire clinical duration.

2. The gas/vapour exchange with the environment should not be hampered because it could cause thermal dropping. This means that the permeability of oxygen and nitrogen (ASTM D1434) and the permeability to the vapour (ASTM E96) should be adequate.

3. The material should not cause an allergic reaction. A composite mixture of elastomeric materials designed for health care application (e.g., an orthopaedic prosthesis) should not elicit an allergic reaction if it is to be introduced onto the market.

4. The heat exchange between the skin and the odoscope sensing head should be minimised. In this case the device was made of Teflon. The package TO92 itself was non-conductive.

5. The material should be easily modelled. A thermally modelled composite elastomeric mixture was acceptable for use. Different materials were tested. We used an elastomeric composite that was 7×10-3 m thick and

thermally modelled. The support for the device can be rigid or flexible depending on the clinical application and

the skin area. It should be observed that the sponge modelled a channel-net, which surrounds each sensor absorbing the sweat. This also avoided sweat flow across the array.

Each lm335 sensor occupies 73.4% of the pixel area, and the absorbing channel-net occupies 26.6% of the pixel area. The effect of this approximation allowed the entire pixel area to be measured through interpolation. The approximation was globally estimated with all of the other error sources.

1.4.2. Modification of the package at the interface

The lm335 package (TO-92) was further modified to minimise the skin-contact area and to furnish geometrical references for interpolation algorithms.

The package was modified at the mechanical laboratory of the Department of Health and Technology to obtain a conical shape at the contact extreme.

A Mod M2 Rosa (Italy) milling machine controlled by the 810 M Siemens computerised system (Germany) was used. The manufacturing limit angle, θ, was equal to 14°. The actual angle, α, used was 10°.

Figure 6 shows a micro-photography of the sensor lm335 before and after the modification. Figure 7 exemplifies a representation of the sensor head with the modified package.


a b

Figure 6. The sensor lm335 before (a) and after (b) the modification

SpongeSensorSupport

SpongeSensorSupport

Sensor A) T0-92 Normal Sensor B) Conic profile angle α = 10°, max angle limit θ = 14°

θαθ

αθ

α

Figure 7. Sensor head and modified package

8


9

1.4.3. Interface material

The material analysis was performed in order to optimize all of the necessary properties of the interface material. Different orthopaedic materials have been unsuccessfully investigated.

In particular, three different elastomeric medical materials were compared from the same manufacturer: Chlorosil ETR 85P11, ETR 85P21, ETR 85P31 (OTTO-Bock, Italy).

Each one was supplied in two parts (A and B), which can be mixed and thermally modelled. They were based on copolymers of methyl-vinyl and dimethyl components of siloxane, which contained a part of silicon for structure. These materials have been used for skin prosthesis interfaces and contained all of the desirable properties of the ideal interface material previously described. We chose the ETR 85P11 material. It had a lower amount of silicon in the mixture, as shown in the ASTM D 2240 test. It had a lower value of hardness when compared to the ETR 85P53. The material ratio was 0.32. This parameter played an important role in the fixation of the device to the skin area. The higher the hardness of the material, the greater the pressure required for contact. The chosen material also showed better transpiration properties for gases (nitrogen and oxygen) during the ASTM D1434 test (ratio=4). It also had better permeability to vapour. In fact, the ASTM E96 (T=25°C) test showed a permeability greater than 70 g/m2/24h while the ETR 85P31 showed a permeability on the order of 40 g/m2/24h.

Figure 8 shows the odoscope with the interface material modelled.

LM335

a b

c

lm335

Figure 8. Images of the odoscope after the material modelling (the lm335 sensors are pointed out)


2. TESTING

2.1. Design of a phantom for the bench testing

There is the need of the development of a dedicated phantom for dynamic bench testing of the device. Equipment designed with this aim could be based on a thermostatic or a Black Body (BB) object. A careful analysis of the market showed that the most frequently used systems are represented by ovens, such as the challenge CH160 (Angelantoni, Italy). This was designed to test and trace the thermal characteristic of commercial components. It showed a high level of stability over long times and had been proven to be useful for static calibration and testing thermal stability of a wearable device. The dynamic bench testing required a system capable of highly precise and controllable thermal waveforms. The starting point was the use of a BB model that changed the temperature according to the driving parameters. However, three main problems arose:

1. A single BB should be large enough to cover the area of the device sensor head. However, this made it very difficult to achieve uniform warming of the area and accurate thermal waveforms to all the sensors of the sensor head.

2. A very accurate system should be used to monitor the thermal waveform generated in a closed loop, real-time system. The monitored values should be used to correct the error between the theoretical and the generated thermal waveform in each pixel area associated with each lm335 component. The difficulty of obtaining this was evident.

3. The range of thermal variation available should be very high (21) to apply to the most frequently used applications of interest.

The phantom was based on 16 small BB units designed with a 10-4 m thin layer of black conductive gum. A precise vertex array for thermal measurements was used in closed loop system to correct the error between the theoretical and the actual temperature produced in real time. This methodology proved to be adequate for bench testing the device. The difference between the generated and theoretical waveforms was never greater than an average of 0.85×10-2°C.

To overcome the three principal problems listed above, we considered that each pixel area monitored by a lm335 component should be bench tested individually by means of a single and properly designed BB. This meant that in the case of the device with 4×4 lm335, the phantom should be composed of 16 independent channels imposing contemporary, but separate temperatures to the 16 corresponding BBs. The BB could be resistive with a thin layer of black conductive gum as well. The driving could be then represented by the current variation and injection. This could solve the first problem previously listed. Owing to the thin layer, the thermal variation was expected to be the same on both sides of the conductive gum. One side could be used to induce the waveform. The other side could be monitored by an array of Precise Vertex Thermometers (PVT). The use of the thin black-resistive conductive gum with the PVT could then solve the second problem. Software procedures could be used to establish current waveforms and produce the waveforms to permit warming of each resistance. The accuracy of such generated waveform (third problem) was the core aim of this work.

Figure 9 describes the phantom in detail and the sensor head of the device. The core of the phantom was the 16 BB units. They were made of a thin layer of black conductive gum with the following characteristics: 30KΩ×10-4m resistance, 25% force deflation (116 N), 125 kN/m2 tensile strength, 109% elongation, 10-4 m thickness, and relative humidity of 50% at 72°C. Each

10


11

BB unit was provided with two thin-foil (10-5 m) copper contacts and had an area of 1.5×10-5 m2. Each one of the units was laterally insulated with Chlorosil ETR 85P11 (OTTO-Bock, Italy), a silicon-based material used in orthopaedics for its compatibility with CE certification. This was also tested as sponge material for the device (6). Figure 9 also shows that the phantom was designed to be positioned on the device for bench testing (1, 6). The BB units permit the production of the thermal waveforms on the lm335 sensor via contact, which monitored each sensor pixel. The other side of the conductive gum-based resistive layer was put into contact with the PVT.

Figure 9. Phantom, in detail (a), and sensor head (b) of the device

(b) SENSOR HEAD

(a) PHANTOM

Thin copperpadInsulating

materialConductivegumThermal

imposition

Spongeof ChlorosilETR 85P11

Lm335 sensor

Elementof the vertexarray

The signals from the PVT system were used to correct the error between the actual generated waveforms and the theoretical waveforms. Figure 10 shows the adiabatic case with both the phantom and the sensor head inside.

Figure 10. The adiabatic case with the phantom and sensor inside


Figure 11 shows the flow chart of the phantom-based dynamic bench testing system.

Measuredtemperature

Control

WEPH

Controlunit

PC

1. Processing of the acquisition

2. Selection of the waveform to be imposed

PROCESSING

Figure 11. Flow chart of all of the equipment for the dynamic bench testing of the wearable sensor (WE) using the phantom (PH)

The simulation environment, fully described in the study of Giansanti et al. (7), was used for the optimization. It showed that if δ (°C) was the mean difference, and if α (°C) was the actual mean temperature, as assessed in ten consecutive samples by the IR system, the optimal value to use for corrections should be:

γ(°C) = ± (0.48+0.98-2α (°C) + 0.87-1α (°C)-2 ) × | δ (°C)|

“+” if δ (°C) <0; “- “if δ (°C) >0. The comparison between waveforms and current injections was performed with tuning

tables. These tables were obtained through several bench test recordings, which account for the actual temperature produced as assessed by the PVT system, the correction value γ(°C) and the produced thermal waveform. Figure 12 shows the architecture of one warming unit channel of the control.

TuningTable

D/A

REG

ITen(K)ITen(t)

AT(K)

SEL(K)

µCPIC16F877

Tuningtable

D/A

REG

ITen(K)ITen(t)

Imposed temperatureto the BB unit by meansof the tensionand then current modulation

AT(K)

SEL(K)

Figure 12. Architecture of a single channel for the control of the warming unit

12


13

The inputs at the time sample K were the digital value of actual temperature AT(K) as assessed by the PVT system, the selected waveform SEL(K), and the digital value of the produced tension as related to the theoretical temperature ITen(K). The first two signals were from the PC. The μP Pic PIC16F877 (Microchip, USA) was used for the communication with the PC, the communication to the Tuning Table and the determination of the correction value γ(°C). All the digital signals were 8 bit. The μP allowed a differential quantization for improving the content of information. The Tuning Table was implemented with a 4-GB compact flash type II memory card (Itachy, Japan).

2.1.1. Generated waveforms

The system designed for the bench testing allowed the generation of polynomial theoretical cubic functions in the form:

fI= aI ±bI× t(s) ±cI×t(s)2 ± dI×t(s)3 It was well known that a monotone increasing or decreasing function may be interpolated by

means of a spline interpolation, which uses cubic functions (22). Furthermore, the spline interpolation proved to be an optimal solution for interpolating the human body’s physical processes in Medical Engineering and Medical Physics (24-26). A body’s thermal process may be divided into time segments individuating sequences of monotone increasing or decreasing functions. The coefficients of the (5) to be produced for each one of the functions were individualized by Matlab R12 spline-based interpolation (23). It was therefore possible to completely reproduce the process.

2.2. Static bench test

The characteristics of the 16 channels were: – cross-talk absent – non-linearity < +- 0.12% fs – hysteresis < 0.09% fs – accuracy <0.3% fs – overall resolution better than 0.03°C in the windows of interest Through a controlled oven and one channel of the PVT, it was also possible to test the

stability of the measurement, which was proven to be better than 0.06°K / 6h in the range determined during calibration. The absence of drifts and the feasibility the equipment for the long time monitoring was shown.

2.3. Phantom-based bench test

Two families of functions produced by the 16 BB units of the phantom were chosen for validation. The choice of these was made on the basis the thermal variation rate and the gradient variation sign to be produced. For the first aspect, the produced waveforms were chosen according to values in the literature. In particular, we took into account medical applications designed with a large range of thermal variation of approximately 8°C / 15 min. The skin was subjected to some unnatural factors (21). For the second aspect, we considered two couples of waves. Each couple realised one increasing waveform starting from the lower value in the range


(25°C), and one decreasing waveform starting from the higher value in the range (41°C). The first couple of waveforms was linear and is described as:

f1-L(t) = a1-L +bL×t + n1-L(t) ( 0 < t < T1)

f2-L (t) = a2-L-bL×t + n2-L(t) ( 0 < t < T2 )

where a1-L= 25°C, a2-L = 41°C, n1-L and (t) = n2-L (t) ≈ 0.01°C are the random error terms. T1 and T2 were fixed at 1800 s. bal was varied in the range +6.9×10- 3 ÷ + 7.1×10-3. The second couple of waveforms was cubic and is described as:

f1-C (t)= a1-C + bC× t(s)+ cC×t(s)2+dC×t(s)3 + nC (t) ( 0 < t < T1)

f2-C (t)= a2-C - bC× t(s) - cC×t(s)2 - dC×t(s)3 + nC (t) ( 0 < t < T2)

where a1-C= 25°C, a2-C = 41°C, and no (t) ≈ 0.01°C are the random error terms. T1 and T2 were fixed at 1800 s, bC was varied in the range +1.0×10-3 ÷ + 1.03×10-3, cC was varied in the range +3.9×10-6 ÷ + 4.0×10-6, and dC was varied in the range +3.3×10-10 ÷ + 3.4×10-10.

Table 1 shows the performance of the dynamic bench test: mean value, maximum value and the standard deviation of the error. These were obtained by comparing the generated waveforms as measured by the PVT system and the theoretical waveforms. The four functions showed similar performances with an error less than 0.98×10-2°C. The statistical significance was very high as shown by Wilcoxon and ANOVA (Analysis Of Variance) (27).

Table 1. Performance of the phantom-based bench tests: mean value, maximum value and standard deviation of the error, obtained by comparing the actual waveforms as measured by the IR system and the theoretical waveforms

F(t) Maximal error °C

Mean error °C

Standard deviation °C

Significance (ANOVA)

Significance (Wilcoxon)

f1-L 1.48×10-2 0.94×10-2 0.32×10-2 p < 0.35×10-3 . p < 0.34×10-3 . f2-L 1.49×10-2 0.95×10-2 0.33×10-2 p < 0.35×10-3 . p < 0.34×10-3 . f1-C 1.52×10-2 0.97×10-2 0.35×10-2 p < 0.38×10-3 . p < 0.37×10-3 . f2-C 1.52×10-2 0.98×10-2 0.34×10-2 p < 0.38×10-3 . p < 0.38×10-3 .

14


15

3. DESIGN OF THE CLINICAL VALIDATION

We followed a predefined thermographic protocol to test the system with 5 healthy subjects. We focused on the temperature variation of the hand metacarpal area during physical exercise (9). In fact, physical activity permitted measurement of very large variations.

3.1. Protocol

The Protocol of the medical validation was based on a motion task to performed using a tapis-roulant according the following instructions:

– Task duration The temperature variation was measured during 1.800 s of running on a commercial tapis-roulant.

– Intensity of exercise The intensity was tuned to 80% of the anaerobic threshold, as estimated by means of a Polar (Polar, Finland) cardiofrequenzimeter.

– Number of trials Each subject performed two batteries of trials (with the wearable system WE), and three batteries of trials with the PVT on the same day.

– Number of subjects Five subjects have been studied.

– Environmental characteristics The environmental conditions were controlled with the temperature at 20°C and 58% humidity.

3.2. Outcome from the clinical validation

Table 2 shows the outcome of the clinical validation. Results indicated that the difference between the reference system (PVT) and the wearable system were very low if the area, temperature or the single pixel area were taken into account. The difference between the two was never greater than an average of 0.09°C.

Table 2. Tempeture error (°C) obtained from the comparison between PVT (reference system) and wearable system)

Comparison Maximal value Mean Standard deviation

Pixel-to-Pixel 0.17 0.08 0.06 Area-to-Area 0.16 0.09 0.05

Figure 13 details the region of affixation according to the clinical protocol.


Integrated thermometer affixation area

Figure 13. Area of affixation according to the clinical protocol

3.2.1. Display of clinical data

A basic issue for a measurement device is the way of displaying data. For this specific system three ways to display data in real-time have been implemented using LabVIEW (National Instruments, USA):

– Mode 1 It shows the 16 signals as functions of time (contemporary or alone) (Figure 14).

Figure 14. Example of 16 channels data-display-interface

16


17

– Mode 2 The second shows the parameters relevant to the pixel temperature (TPixel) and to the area temperature (TArea).

– Mode 3 The third (considering that the wearable system is also an image transducer) is a two dimensional colour-mapped display during the clinical application (Figure 15).

Figure 15. Example of the two-dimensional map of the interface


4. IN DEPTH ANALYSIS

4.1. Details on the device positioning

The sensor head can be developed with a rigid support or flexible one. The latter permits the complete adaptation of the sensor head to a curvy and/or rough skin region. Furthermore, the ETR 85P11 material had a low hardness, as shown by the ASTM D 2240 test. This also simplified the positioning problems. The sensor head can be affixed with very low pressure, which can be improved with an elastic bandage. The precise contact can be checked in real time. An acoustic alarm switches on when problems occur.

4.2. Assuring aseptic conditions

The materials used in the sensing head were inert. Furthermore, the service unit did not have UV erasable components.

Two solutions emerged during this preliminary analysis: – The sensing head could be sterilised with UV radiation before each trial (a fluid-based

sterilisation treatment was also possible). – The sensing head could be left with the patient for the entire period of therapy and

sterilised by the patient changing by means of the UV radiation.

4.3. Device certification

The device has been developed for our use and will not be destined for the free market. We will produce only a few prototypes for our exclusive clinical research. However, we respected the development requirements for the certification CE of Medical Devices (European directive 93/42/CEE on medical devices, actuated in Italy by means of the Italian legislative decree n. 46 dated 24-02-97) and ensured the patient was free from electrical shocks, free from incompatible interaction between skin and the sensing-head, and guaranteed overall patient protection.

4.4. Device life, costs and maintenance

The sensor unit was a prototype. The estimated future cost of product production for one exemplar is 450 Euros (Table 3).

To extend the life of the device we used components with military specific failure rates and conducted an observation study of the critical components to complete large scale production. A preliminary conservative stress-analysis study showed 18,000 hours of life for the sensor. We also estimated that the cost of labour for the replacement of an integrated component with Surface Montage Devices (SMD) package was less than 3 Euros. This was due to the very short average time for substitution (10 minutes) thanks to the use of MANTIX VISOR (Vision Engineering, USA), which is specific for SMD montage.

18


19

Table 3. Cost of the system in details

Description Tipology N. Cost (Euro)

Microcontroller PIC Pic 16F877-04/pt (Microchip, USA) 1 15.44 Temperature sensors lm335 (national, USA) 16 16.32 Transmitter and receiver 8110-xtr (Aur°el, USA) 1 38.00 Development of the masters in surface montage technology

1 50.00

Construction of the case 1 30.00 Silicon material at the interface 1 20.00 Power supply 1 44.73 Montage 20 hours of work 1 200.00 Further electronic components - 35.00

Total 449.49


5. DISCUSSION AND CONCLUSION

The system described permits the thermographic continuous monitoring and reconstruction of the skin region to which it was affixed. Prior to this paper, the problem of dynamic thermal continuous monitoring with contact technologies was offered by Bolton (3). The previous solution permitted the contemporary thermal contact monitoring of different height points and was intended for a specific medical application where it was necessary to monitor different body points.

Our device was designed for the monitoring of a defined skin area. It permitted the continuous thermographic monitoring with high precision, a complete colour mapping reconstruction and minimisation of the typical problems of contact thermography. Furthermore, the system can be furnished with a different sensor and number of pixels (16 to 256 pixels are available) (4, 8). The measurement was direct and was not influenced by the environment, which could happen for IR telethermography. The measurement was not hampered by the geometry of the skin surface (curvatures, roughness) thanks to the flexible adaptation of the sensing head to the surface. In the IR telethermography, curvatures generated aberrations in the radiation. Furthermore, the prototype cost of production for one exemplar was 20 times lower than the costs of an IR system with a same device-life. The device life was tested with a stress analysis. The clinical test conducted on a percentage of healthy subjects using a PVT comparison and by choosing an application in the range of the device showed the feasibility of the developed device. In fact, the maximum difference between the calculated and real data, as assessed and recorded by the two systems over the entire area, was lower than 0.16°C with a standard deviation lower than 0.05°C. The greatest difference, as assessed per pixel, was lower than 0.17°C with a standard deviation lower than 0.06°C. Considering the peculiarity of the chosen clinical application (a wide range of thermal changes over a varying amount of time, sweating, etc.), the system could be successfully used in for any application of skin thermography (11-18), which was also detailed in the previous paper (8). Furthermore, the measurement over long time periods could permit the recovery of the inheritance of some studies conducted in chronobiology such as those described in (5-6).

The use of the device could permit the development of medical knowledge with protocols utilising one side of the instrumental assessment for direct patient interaction. The other side could be used for a comparison of the assessed data and expected data forecasted with suitable bio-heat-equations that were properly designed and specific for the medical applications (8, 19, 20).

The spatial resolution depended on the use of the traditional package, TO92. To improve this design parameter we can follow two directions. The first direction was the package reduction and the use of the Surface Montage Technology. The second direction was the use of particular procedures and interpolation algorithms. The investigation on package reduction did not furnish results. In fact, the same manufacturer recently planned the production of the unpackaged die and wafer version of the lm335 component with the identical data-sheet properties. The use of the unpackaged die could improve the spatial resolution by a factor of 22 and 24. Furthermore, this unpackaged die could be easily positioned on the elastomeric material. The production of the unpackaged lm335 die version was discontinued and could be justified only for large industrial purchases, such as military ones. However, its use was feasible, attractive and gives full justification to the current study.

Many manufacturers have yet to investigate the thermal non-contact technologies that use interpolation procedures and algorithms. Skin thermal gradient variations between continuous

20


21

areas are not easily calculated, and interpolation could be a solution. There are interesting solutions such as the IRISYS (Milton, Keynes, UK), a no-contact thermal sensor, 16X16 output, which can be interpolated by a software procedure to obtain a 128×128 pixel image.


REFERENCES

1. Bonato P. Wearable sensors/systems and their impact on biomedical engineering. IEEE Eng Med Biol 2003;22:18-20.

2. Bolton P, Mc Guinnes K, Cooley J, Hayes L, Howell A. A compact data logger for ambulatory skin temperature measurement. Journal of Med Eng Tech 2001;25(6):264-8.

3. Giansanti D, Maccioni G, Gigante GE. The project of a wearable device for mobile skin thermography. In: Congress on Medicon and Helath Telematics 2004. Ischia, 31 luglio-5 agosto 2004, IFMBE Proceedings 2004;6: Poster 477.

4. Simpsons HW, Griffits K. The diagnosis of breast pre-cancer by the chronobra -I. Background review. Chronobiology Intemational 1989;6:355-69.

5. Simpsons HW, Griffits K. The diagnosis of breast pre-cancer by the chronobra -I. The breast precancer test. Chronobiology Intemational 1989 6:371-93.

6. Linear data book. Manual 1. New York: National Semiconductor Corporation USA; 2002.

7. Giansanti D, Maccioni G, Gigante GE. A comparative study for the development of a thermal odoscope for the wearable dynamic thermography monitoring. Med Eng Phys 2006;28(4):363-71.

8. Giansanti D, Maccioni G. Development and testing of a wearable integrated thermometer sensor for skin contact thermography. Med Eng Phys 2007;29(5):556-65.

9. Zontak A. Dynamic thermography: analysis of hand temperature during exercise. Ann Biomed Eng 1998;26:988-93.

10. Linear data book. Manual 2. New York: National Semiconductor Corporation USA; 2002.

11. Ring E, Watson C, Barker J. Infrared thermography and IC thermal clearance of the skin. Applied Thermology 1985;5:133-41.

12. Ambar M. Diagnostic thermal imaging: a historical technological perspective. In: Ambar M (Ed.). Quantitative dynamic telethermography in medical diagnosis. CRC Press: Boca Raton 1994. p. 333-40.

13. Muth H, Wormald P, Bishop J, Donald K. Further studies of blood flow in the resting arm during supine leg exercise. Clin Sci 1958;17:603-12.

14. Kirsis F. Issledovanie kozhnogo krovtka v nerabutaushkey konechnosti pri fizicheskoi nagruzke. Teor Prackt Fiz kul 1975:8:29-31,

15. Johnson J. Exercise and the cutaneus circulation. Exercise sport Sci Rev 1992;20:59-97.

16. Sterns EE, Zee B, SenGupta S, Saunders FW. Thermography. Its relation to pathologic characteristics, vascularity, proliferation rate, and survival of patients with invasive ductal carcinoma of the breast. Cancer 1996;77(7):1324-8.

17. Kurihara T, Higashi Y, Suemasu K, Kanoh T, Tabei T, Inoue K. Usefulness of contact thermography for the evaluation of chemotherapeutic effectiveness in breast cancer. Gan To Kagaku Ryoho 1993;20(7):915-20.

18. Sforza M, Ballerini A, Russo R, Carzaniga PL, Vertemati G. Contact thermography in breast pathology. A critical review. Minerva Chir 1991;46(8):375-7.

19. Sanczyk A, Telega JJ. Modelling of heat transfer in biomechancs- a review Part II . Orthopaedics Acta of Bioengineering and Biomechanics. 2002;4(2):409-18.

20. Suktanski M, Yablonsky JJ. An analytical model of temperature regulation in human head. Journal of thermal biology. 2004; 29:554-59.

22


23

21. Vainer BG. Treated skin temperature regularities revealed by IR thermography. In: Inchingolo P. Proceedings of SPIE, Trieste 16-18 settembre 2004. Trieste: EUT, Università di Trieste, 2004. p. 470-481

22. Hairer E, Norset SP, Wanner G. Solving ordinary differential equations I: nonstiff problems. New York: Springer-Verlag; 1991. (Springer Series in Computational Mathematics vol. 8.)

23. Matlab software version 3. Manual. Natick, MA: The Mathworks Inc.; 2003.

24. Meijering EH, Niessen WJ,.Viergever MA. Quantitative evaluation of convolution-based methods for medical image interpolation. Med Image Anal 2001;5(2):111-26.

25. Lehmann TM, Gonner C, Spitzer K, Addendum H B-spline interpolation in medical image processing. IEEE Trans Med Imaging 2001;28(5):660-5.

26. Thevenaz P, Blu T, Unser M. Interpolation revisited. IEEE Trans Med Imaging 2000;19(7):739-58.

27. STATA software manual. College Station, TX: Stata Corporation USA: 2003.


24


25

APPENDIX Technical details

of used component and circuits


26


27

The Appendix contains the optimal calibration curve of the lm335, the conditioning chain with the Sallen & Key filter, and the polarisation circuitry of the lm335 in the Celsius thermometer configuration.

A.1. Optimal calibrated curve of the sensor lm335

The component lm335 (National, USA) has little calibrated error in the range of interest of the human body skin temperature. Figure A1 shows this error as a function of the temperature.

Error (°C)

-0,5

0

0,5

1

1,5

2

-55 -15 25 55 105 145 185

Temperature ( °C)

Figure A1. Optimal calibrated curve of the lm335 in the temperature range of interest

A.2. Signal conditioning with the Sallen & Key cell

Figure A2 illustrates the Sallen & Key cell used as conditioning circuit. In order to obtain the Butterworth approximation, Q=1/√2, Rb/Ra was set at 0.6, the Gain G2 =1+0.6, and G1= 52.125.

-

+

15V

Vin100pF

R2

lm308

R1

100pF

R R

100pF

15V

Vin

R2

R1

R R

100pF

lm308

-

+

-

+

15V

Ra

Rb

Ra

VCC

precise double potentiometerfor the frequency swap

Figure A2. Sallen & Key cell with the optimized parameters


A.3. Polarisation of the lm335 component

Figure A3 shows the complex polarisation circuitry for the lm335 component in the Celsius thermometer configuration with a potentiometer and a trimmer.

10 mV / °C

-

+

VC

lm336 100p

1K

VC

lm308

6K

--

VCC

VCC

9K

lm335

F

12 K

1K

VCC

6K

Figure A3. Polarisation of the lm335

28

La riproduzione parziale o totale dei Rapporti e Congressi ISTISAN deve essere preventivamente autorizzata.

Le richieste possono essere inviate a: [email protected].

Stampato da Tipografia Facciotti srl Vicolo Pian Due Torri 74, 00146 Roma

Roma, luglio-settembre 2008 (n. 3) 10° Suppl.

ISTITUTO SUPERIORE DI SANITÀold.iss.it/binary/publ/cont/08-24 web.1222954103.pdf · vascolari,...

Documents

Transcript of ISTITUTO SUPERIORE DI SANITÀold.iss.it/binary/publ/cont/08-24 web.1222954103.pdf · vascolari,...